基于模糊小波神经网络的BDI模型

BDI Model Based on Fuzzy Wavelet Neural Network

为克服原有BDI模型可计算性差以及不能处理模糊问题的弱点,提出一种基于模糊小波神经网络(FWNN)的BDI模型,FWNN用神经网络来实现模糊化、模糊承诺和去模糊化的过程,并利用小波基函数作为模糊隶属函数,网络权值和隶属函数的形状均是可学习调整的。以一对一追逃问题为背景的仿真实验验证了模型及算法的可行性。

The BDI model is not efficiently computable and not easily applicable to the domain with fuzzy. A BDI model based on fuzzy wavelet neural network （FWNN） was proposed. FWNN employed neural network to implement fuzzification, fuzzy commitment, and defuzzification, and used wavelet-based function as member function. So both the network weights and the shape of member function could be learned. The experiment with one to one agent pursuit-evasion game shows that the proposed model and its algorithm are feasible.